It is common practice to use classification techniques that use character information, such as music titles added to the music, composers, and artists, as a method for categorizing music into a list or the like.
For example, in a disclosed method for categorizing music, analog sound that includes multiple pieces of music is converted to digital sound and the start and the end of each piece of music are specified. Subsequently, character recognition is performed on character information (meta information) related to the music in which the start and the end are specified. Then, the music in which the start and the end are specified is associated with the result of the character recognition.
Furthermore, in another disclosed method for categorizing music into multiple pieces, items, such as titles, artist names, genres, review texts, tempos, beats, and rhythm, are extracted from meta information contained in the music. Then, each of the extracted items is associated with the music from which it was extracted.
Patent Document 1: Japanese National Publication of International Patent Application No. 2006-515099
Patent Document 2: Japanese Laid-open Patent Publication No. 2007-26425
However, with the conventional technologies, music is classified using character information that is added to music; therefore, there is a problem in that, if the amount of character information added to the music is small, the music is not accurately classified.
Furthermore, if classification is performed using musical characteristics, such as tempos or tone, without using information added to the music, the classification becomes overly specialized, and thus it is not a general purpose classification that can be used by general users. Furthermore, there may also be a method for classifying music after character information is added to the music; however, character information needs to be created and then added to each piece of music, which takes a lot of time and effort, and thus it is impractical.